test: Felt constructor edge cases

[mooori]

Closes #931

[greenhat]

Nice! I'm sorry I should have articulated it more, but after our "wrapping nature of the felt" discussion on slack I made the Felt::new on miden target in the miden-crypto to behave like the native felt (wrapping). So in the test we should check that Felt:new compiled to the miden target is semantically equivalent to the one compiled to the default target. However, this will be possible after we migrate to the VM v0.21 with the "unified" Felt from the miden-crypto repo.

[bobbinth]

Felt::new on miden target in the miden-crypto to behave like the native felt (wrapping)

After the migration to Plonky3, this is not longer the case - i.e., Felt::new() does not reduce the passed-in value to a canonical value. Though, I think Felt::new(x).to_canonical_u64() would give us the wrapped value pack (I haven't double-checked this though).

[greenhat]

Felt::new on miden target in the miden-crypto to behave like the native felt (wrapping)

After the migration to Plonky3, this is not longer the case - i.e., Felt::new() does not reduce the passed-in value to a canonical value. Though, I think Felt::new(x).to_canonical_u64() would give us the wrapped value pack (I haven't double-checked this though).

Yes. Thank you! This is what I was trying to say. I meant "non-reduced" instead of wrapped. It becomes wrapped after the "roundtrip". My bad.